Class reading optimization [1/4]: optimize CharacterArray class>>decodeFromJavaUTF8: This method now uses hand-optimized inline C to decode bytes. Also now it takes instance of a String instead of a ByteArray to avoid excessive byte copying from byte array to string in case all characters are below 0x80. An Unicode16String is create only when input bytes encode a character greater then 0x80. *** Classr reader optimization [3/x]: call self info: conditionally if Verbose flag is set. This avoids an extra message send and a lots of string concatenations. Thanks Claus Gittinger for spotting this.

Added JavaSourceFileWriter>>fileOut:on:withTimestamp:... Used by SCM to write class source to a file.

Merged 945bee015640 and b6bced0551a9

Merged d9a3d685c8b9 and 5260fddf10f6

Merged 304d3543242f and 603fdba602c4

Fixed annotation support in reflective environment (getValue() and standard tag bits) Changed ReflectiveElementValuePair#getValue() to actually return values as specified in IBinaryElementValuePair. This fixes compiler when compiling annotated classes. Also fixed standardTagBits() to be more in sync with ClassReader as it seems that Eclipse Compiler relies solely on them.

Cleanup in ReflectiveEnvironment: more comments & logging. Added more comments here and there. Use java.util.logging when .jar could not be opened.

Force a garbage collect after each run of classreader benchmark. Classes read are garbage anyway and this avoids GC routines to garble profiling results.

Fixed JavaRelease>>validateJavaVersionString: Allow for extra content before 'java version..." string. There may be some (garbage) output from linker or whatever (actually happens when profiling libjava under callgrind).

Made JavaClass>>package returning Smalltalk/X package name from annotation, if annotated. The stx.libjava.annotation.Package marks a class to belong to a Smalltalk/X package so IDE tooling, especially source code managers, can use this information (and commit Java classes along with Smalltalk classes, for instance)